We built a real native app on both platforms in a fraction of the time it used to take. It’s called Phloat. We originally designed it to solve the “Your phone is full” problem that Apple and Google ask ~$10/mo to solve for you.
Prototype in hours, not weeks
With Claude in the loop we had a working prototype in a day. It ran on a real iOS device with a working picker, a real upload path, and thumbnails coming back from storage. The Android version followed in days, written as a proper native Kotlin app rather than a port, because the cost of native Android versus a bridging solution like React Native was also zero.
The reason it works is that the expensive part of native development was never the typing. It was the hundred small platform decisions in between: how PHPickerViewController hands you an asset, why EXIF rotation and MediaMetadataRetriever disagree, what a 12-megapixel decode does to your memory ceiling when a pager is holding three of them — and having engineers who know those details well enough to steer them in the right direction.
Those are the things that used to turn a two-day feature into a two-week feature.
Iteration at the speed of ideas
Phloat started as an offloading tool. Free space safely and cheaply. When people started using it, the thing they kept doing was not offloading but making albums. So we built shared albums across devices, which meant album capture, join-by-QR, multi-contributor libraries, abuse controls — a whole second product surface.
In a traditional build that pivot is a hard conversation. You are months in, the architecture has opinions, and changing the premise means renegotiating scope, timeline, and budget all at once. What actually happened was that the pivot took about as long as deciding to make it.
AI is a tool, not a replacement for judgment
Everything good about Phloat came from a human decision.
Choosing to count only originals against someone’s quota, and never bill them for thumbnails. Not loading full-resolution originals over a metered connection, because those bytes belong to the person paying for them, not to us. Using the system’s own navigation bars and search fields instead of drawing convincing fakes, after a redesign taught us exactly what the fakes cost. Noticing that the people in the most pain were not the ones comparing cloud storage prices, but the ones who had already given up and were deleting memories to get by.
Claude did not have those ideas. Claude made them cheap to act on, which is a different and genuinely valuable thing. Taste, product judgment, and knowing which problem is worth solving are still entirely ours, and they are still the whole ballgame. The tool just removed the tax between having a good idea and seeing whether it was one.
What’s next
This is now how we work, and it’s available to the people we work for.
Northwoods builds custom native apps, iOS and Android, in weeks rather than months. Real Swift and real Kotlin, running on the platform’s own controls.
We have two customer projects in flight right now that we are genuinely excited about, in very different industries, both the kind of thing that would have been quoted as a six-month engagement a year ago. More on those soon.
If you have an app that has been sitting in the “someday, when we have the budget” column, the math on that may have changed. Get in touch, or go try Phloat and see what a few weeks looks like.